Some thoughts on human + agent collaboration:
1. Slack has become the default place where humans work with agents and each other. Yet many people I talk to don't actually enjoy using it. Monitoring dozens of channels and DMs is overwhelming.
2. You can of course have your agent monitor all those threads for you, but that creates a new problem. I recently gave feedback to someone at an AI lab and got a quick reply. But it turns out it was their agent that responded, not them. If agents keep multiplying in Slack channels, will it only get harder to reach the actual human? And if only the agent replies, how do I know my feedback is getting prioritized?
3. Meanwhile, Grok Bot, ChatGPT, and others are starting to look more and more like Slack as they expand from human-agent to human-human collaboration. But going back to my first point, Slack is already overwhelming to use.
4. Maybe the solution is an agent that handles 80% of your threads and escalates the other 20% to you. But I doubt it'll pick the right 20% most of the time. I personally don't like it if I know something I manually wrote is getting filtered by an agent instead of going straight to a human.
There is only One True SaaS Layout.
You may not like it, but this is what peak usability looks like